The plot on the left shows how a classifier tries to distinguish between the music from Limburg and Brabant. What becomes clear is that the classifier is better at telling if the music from Limburg is Limburgs or Brabants. It got 83% of the songs right. Vice versa, the classifier performed worse. For Brabant it got only 63% right. I honestly have no clue why the difference is this big.
In this storyboard I’m going to look at Carnival Music, in Dutch this is called “carnavals muziek.” At first, I’m going to investigate how it differs from modern hitsongs. I think this is very interesting, because it is both music that people easily can sing along to. Also, when you listen to it, you can clearly hear how different each kind of music is, but it is both very popular.
After analysing the difference, I am going to take a look at the difference in carnival music itself. Because it is celebrated in two different provinces in The Netherlands there must be a kind of difference in the music too. One main difference will be the language. This is due to the fact that the lyrics of the music in Brabant are in Dutch and in Limburg they are in Limburgs, which can be compared to a mixture of Flemish, Dutch and German.
I will do this by using different Spotify playlists that make up my corpus. Also, I think that comparing various audio features to each other can give me an idea about what makes each kind of music different. After doing this for carnival music vs. current hits and Limburg vs. Brabant, I will elaborate further from that point.
What Is Carnival?
Originally, Carnival (in Dutch Carnaval or Vastelaovend) is Catholic feast.[1] It is officially celebrated on the three days before Ash Wednesday, but in practice, everyone mostly starts celebrating on thurday instead of sunday. Ash Wednesday marks the beginning of the Lent and thus for a lot of people the last chance to party. The word vastelaovend in Limburgs means vastenavond in Dutch, which in English can be translated to Lentnight. This, in turn, could be led back to the night before Lent.
The carnival season starts on the 11th of November, 11/11 at 11:11 am. Eleven is seen as the number of the crazy and that is why almost everything has some attachment to this number.
Carnival is celebrated in the south of The Netherlands and mostly in the provinces Brabant and Limburg. The ways it is celebrated can also be diveded in two mean variants; “het Bourgondisch Carnaval” and “het Rijnlands Carnaval.” In Brabant you will see especially the first one. Everyone dresses simple and the same, because everyone is the same. This way, classes are not visible anymore. The second kind of celebration is the one you will see in Limburg. Everyone dresses extraordinary. The crazier the better. Ofcourse, there are a lot of local adaptations of these two main kinds of celebration.
[1] Wettelijke Feestdagen. Carnaval. https://www.wettelijke-feestdagen.nl/Feestdagen/Carnaval.aspx.
Vastelaovesgezicht 2019
What Does This Scatterplot Show?
For this visualisation I used three different Spotify playlists; Vastelaovend 2020, Carnaval Brabant and Nederlandse Hits 2020. To compare the carnival music to the current hits in The Netherlands, I didn’t distinguish between Limburg and Brabant. I compared the valence, on the x axis, to the energy, on the y axis, of both playlists. The size of the dots gives an idea of the loudness of each individual song. In the table below the means and standard deviations of each of Spotify’s audio features is shown. I have chosen valence, energy and loudness because, as you can see, the difference between these features is the highest.
| Audio Feature | Carnavival Music | Hits in The Netherlands |
|---|---|---|
| Acousticness | M = 0.213 SD = 0.173 | M = 0.291 SD = 0.272 |
| Danceability | M = 0.669 SD = 0.135 | M = 0.665 SD = 0.123 |
| Energy | M = 0.841 SD = 0.109 | M = 0.619 SD = 0.163 |
| Instrumentalness | M = 0.008 SD = 0.059 | M = 0.009 SD = 0.067 |
| Liveness | M = 0.250 SD = 0.198 | M = 0.155 SD = 0.121 |
| Loudness | M = -5.44 SD = 1.66 | M = -6.38 SD = 2.45 |
| Speechiness | M = 0.084 SD = 0.065 | M = 0.086 SD = 0.081 |
| Valence | M = 0.792 SD = 0.180 | M = 0.511 SD = 0.220 |
| Mode | M = 0.876 SD = 0.330 | M = 0.630 SD = 0.485 |
Means and Standard Deviations of various Spotify Audio Features
What Is The Difference Between Carnival Music And Current Hitsongs?
What stands out the most is that the carnival music is mostly situated in the same quarter, high energy/high valence, while the hitsongs are more spread out, only avoiding the low energy/high valence quarter. This indicates that carnival music is generally very similar. As is shown in the table and the scatterplot, carnival music is more often in a major mode. The music is almost always “feel-good” music and so, the use of a major mode seems like a logical decision.
What Tracks Are The Most Popular?
To compare the songs from Brabant to those from Limburg, I need to make a selection of which songs I will use. I wanted to base this decision on track popularity.
For Brabant, the three most popular tracks are De Toreador, Links Rechts and Brabant. The first song is definately an outlier and that is why I won’t be using that song. Brabant by Guus Meeuwis is a song that is played by Dutch radiostations throughout the whole year, so I won’t be using that song either. Therefore, the songs I will use are Links Rechts, Kali and Potentie.
Limburg by Rowwen Hèze is the most popular song for Limburg, but the same as for the Guus Meeuwis applies to this song. The songs I will be using are Nao ’t Zuuje, VRUNJ TOT DE ALLERLESTE RUNJ and 1000 Sterre.
Another thing that becomes visible by this plot is that the Dutch music is much more popular. This could be due to the fact that other people, who don’t celebrate carnival, still listen to this music at e.g. parties. Because these songs are in Dutch and the songs from Limburg aren’t, makes these more approachable for others.
For making the tempograms, I used the same songs as before. What becomes clear is that even though the songs seem to be very similar, the tempi of the songs really differ from each other.
The song 1000 Sterre has a tempo a bit slower then 140 BPM. In the pre-chorus of the song, there is no drumbeat and this is very clear to see in the tempogram.
The second song, Nao ’t Zuuje, is a very difficult song for Spotify to define. I think this is due to the fact that the song shifts between two time signatures, 6/8 and 9/8. And the song makes almost no use of drums.
Joost is the song with the most clear tempo around 150 BPM. When you listen to the song, you could argue that it has been made with a computer. Personally, this tempogram is an argument to say that this is indeed what happened.
I think that something went wrong in the last tempogram. Because, according to this, there should be an increase in tempo at the end of De Zuipschuit. I don’t really hear this myself, but that could be something to look into a little bit deeper.
I made four keygrams, two for songs from Limburg and two for Brabant. Though the keygrams don’t show the right one, you can see the different parts of the songs. It becomes clear that, despite the fact that carnival music in general is very similar, individual songs show difference in structure.
Nao ’t Zuuje, for instance, consists of four different parts in which only the last part is different in key. 1000 Sterre also has four parts, but the chorus couplet structure is clearly visible. Both the keygrams show a lot of yellow.
When you look at the keygrams for the songs from Brabant, you see a lot more blue. Joost consists of three parts; the small yellow part, the middle and the modulation at the end of the song. De Zuipschuit shows four parts. I don’t think it is possible to tell in which key this song is made, but in the middle something else is happening. At the end, there are a lot of changes.
Because there is a language difference in the music from Brabant and Limburg, I thought it’d be interesting to compare a song. Sometimes artists make covers of songs that are popular in the other county. In this case I choose the song Laot de zon in dien hart and Laat de zon in je hart, which means let the sun into your heart. When you listen to the tracks, the two versions seem to be very alike.
| Audio Feature | Rene Schuurmans | Beppie Kraft |
|---|---|---|
| Danceability | 0.731 | 0.786 |
| Energy | 0.807 | 0.791 |
| Key | 9 | 3 |
| Loudness | -5.508 | -6.294 |
| Mode | 1 | 1 |
| Speechiness | 0.0289 | 0.0279 |
| Acousticness | 0.636 | 0.431 |
| Instrumentalness | 0 | 0 |
| Liveness | 0.275 | 0.332 |
| Valence | 0.963 | 0.939 |
| Tempo | 123.986 | 123.94 |
As becomes clear from the table, the only big difference between the two songs is the ‘loudness’ feature and the key. The version of Rene Schuurmans is in A major and the version of Beppie Kraft is in D# major.
I think the reason that the graph is so blurry is because the songs are in different keys, but I am not sure and I don’t know how to fix it.
For the self similarity matrices, I choose the song 1000 Sterre by Bjorn & Mieke. It is one of the song I like the most. There is a clear pattern visable and it shows that the song is repetitive. I also choose bars as level of detail, because this gave the most clear matrix.
I wasn’t able to make the chroma self similarity matrix work and because of that I could not compare the timbre on to the chroma one. I hope this is not a problem.